Resources

Use a locked Python environment, PyTorch, tokenizer tooling, profiling, structured run metadata, and a frozen evaluation harness. Primary sources are framework documentation and original research papers.

Evidence and review

Every artifact records model version, tokenizer, precision, hardware, seed policy, and data fingerprint. Claims must trace to versioned code and artifacts. Screenshots, unexplained generated prose, and a single favorable run do not pass.

Oral defense

Explain the mechanism, identify the most dangerous assumption, interpret one failure from raw evidence, and name the condition that would reverse your decision.

Source backbone

Use Dive into Deep Learning, PyTorch documentation, primary papers, and Designing Data-Intensive Applications selectively.